Do species and functional groups differ in acquisition and use of C, N and water under varying atmospheric CO2 and N availability regimes? A field test with 16 grassland species


Author for correspondence: Peter B. Reich Tel: +1 612 624 4270 Fax: +1 612 625 5212


  •  To evaluate whether functional groups have a similar response to global change, the responses to CO2 concentration and N availability of grassland species from several functional groups are reported here.
  •  Sixteen perennial grassland species from four trait-based functional groups (C3 grasses, C4 grasses, non-leguminous forbs, legumes) were grown in field monocultures under ambient or elevated (560 µmol mol−1) CO2 using free-air CO2 enrichment (FACE), in low N (unamended field soil) or high N (field soil +4 g N m−2 years−1) treatments.
  •  There were no CO2 × N interactions. Functional groups responded differently to CO2 and N in terms of biomass, tissue N concentration and soil solution N. Under elevated CO2, forbs, legumes and C3 grasses increased total biomass by 31%, 18%, and 9%, respectively, whereas biomass was reduced in C4-grass monocultures. Two of the four legume species increased biomass and total plant N pools under elevated CO2, probably due to stimulated N-fixation. Only one species markedly shifted the proportional distribution of below- vs aboveground biomass in response to CO2 or N.
  •  Although functional groups varied in responses to CO2 and N, there was also substantial variation in responses among species within groups. These results suggest that current trait-based functional classifications might be useful, but not sufficient, for understanding plant and ecosystem responses to elevated CO2 and N availability.


Generalizing about responses of different plant species to elevated atmospheric CO2 concentrations or N deposition remains an elusive goal in global change biology. Given the wide variety of species, any kind of grouping that simplifies the variation among species while providing predictive power will significantly advance the field. Based on well-known differences among species in important intrinsic traits, a variety of hypotheses have been developed about potential functional group differences in response to CO2 and N enrichment.

1) Intrinsic differences in photosynthetic biochemistry should lead to markedly greater responsiveness to elevated CO2 for C3 than C4 plants (H1). Although supported by theory and early field studies (Pearcy & Ehleringer, 1984; Curtis et al., 1989), more recent studies suggest such differences may be less pronounced (Owensby et al., 1993; Wand et al., 1999), especially during dry periods.

2) Productivity in N-fixing legumes may be stimulated by elevated CO2 more than in nonfixers (H2), because the former should be less N-limited. It has long been known that elevated CO2 stimulates legume growth and N2 fixation (Finn & Brun, 1982; Zanetti et al., 1996). N-fixing species have often shown a stronger biomass response to elevated CO2 than nonfixing species (Soussana & Hartwig, 1996; Clark et al., 1997; Hebeisen et al., 1997; Lüscher & Nösberger, 1997, 1998; Schenk et al., 1997), although little work to date has been on wild species at naturally low levels of N availability.

3, 4) A series of related hypotheses posit that increases in CO2 or N supply should lead to a more pronounced growth increase in species of given strategies, habitats, or growth rates. For our study we propose the hypothesis (H3) that C3 grasses considered more disturbance adapted and nitrophilic, should respond more to increase in N supply than C4 grasses (cf Wedin & Tilman, 1996). Moreover, being less N-limited, the legumes should also be less responsive to N addition than the non-fixers (H4).

5) In addition to understanding the effects of elevated CO2 or N singularly, given the potential for the CO2 fertilization effect to be modulated by N-supply, it is important to study species and functional group responses to combinations of these two elements. Many ecosystem models theorize that CO2 responses are constrained by N limitation (but see Cannell & Thornley, 1998) and actual evidence is mixed (Larigauderie et al., 1988; Owensby et al., 1994; Leadley & Körner, 1996; Lloyd & Farquhar, 1996; Poorter et al., 1996; Volin & Reich, 1996; Zak et al., 2000). If the CO2 response of species is generally N-limited, then nonlegumes, without the ability to fix atmospheric N and therefore modulate their own N supply, should demonstrate smaller responses to elevated CO2 than legumes (see H2) and be more likely to show a CO2 × N interaction (H5).

6) For individually grown plants, usually in their first year, root fraction (root biomass as a fraction of total biomass) adjusted for ontogenetic drift is unaffected by CO2 (Curtis & Wang, 1998; Reich, 2001) and is lower under enhanced N supply (Poorter & Nagel, 2000; Reich, 2001). Data for older plants or stands are rare. If assemblages over multiple years behave similarly as young, individual plants, one might hypothesize that elevated CO2 would have no effect (H6a) and N addition a decreasing effect (H6b) on root fraction in our experiment. Since root fraction incorporates both allocation and turnover, neither of which are well documented in the field (Reich, 2001), these hypotheses are proposed as null models.

7) Effects of treatments on biomass and physiology should be reflected in plot-scale resource availability. Assuming that greater CO2 and N supply both lead to increased biomass, we hypothesize (H7) that soil solution N concentration and percentage soil water should decrease under both treatments. However, increased N supply should compensate for increased N uptake, minimizing the decline in soil N compared with that under elevated CO2 (H7a). Under elevated CO2, reduced leaf level water loss could minimize the decline in percentage soil water compared to that experienced in the high N treatment (H7b).

Although there are an increasing number of tests of the CO2 × N interaction hypothesis in general and of the functional group-related hypotheses raised above, few have been done in field settings where both root and shoot processes can be quantified for more than a small number of species. To help fill this gap, we addressed these issues using an experiment comprising 16 grassland species from four functional groups grown in 128 monoculture plots that is a part of a larger experiment (BioCON) designed to test interactions among species diversity, elevated CO2 and N deposition (Reich et al., 2001). In particular, we assessed whether functional groups, growing in monoculture plots in a free-air CO2 enrichment (FACE) experiment, differed in their acquisition and use of C, N and water in response to combined treatment with elevated CO2 and N deposition.

Materials and Methods

The BioCON (Biodiversity, CO2 and N) experiment (Reich et al., 2001) (;) is located at the Cedar Creek Natural History area, a National Science Foundation, Long-Term Ecological Research site in Minnesota, USA (lat. 45° N, Long. 93° W). The region has a continental climate with cold winters, warm summers (mean January and July temperatures of −11 and 22°C), and precipitation averaging 660 mm y−1. The soils are derived from a glacial outwash sand plain and are sandy and nitrogen poor. Plots were established on a secondary successional grassland after removing the previous vegetation.

Our study included 128 individual monoculture plots (each 2 × 2 m), a subset of all plots, distributed nearly equally among six 20-m diameter experimental areas (rings). In three elevated CO2 rings, a free-air CO2 enrichment (FACE) system (Lewin et al., 1994) was used during the 1998 and 1999 growing seasons to maintain the CO2 concentration at 560 µmol mol−1. Three ambient rings (368 µmol mol−1 CO2) were treated identically but without additional CO2. The experimental treatments were arranged in complete factorial combination of CO2 (ambient or elevated), species (a total of 16, four from each of four functional groups), and N level (low and high) for a 2 × 16 × 2 design with two replicates. Each plot was planted in 1997 with 12 g m−2 of seed. The design consisted of a split-plot arrangement of treatments in a completely randomized design. CO2 treatment is the whole-plot factor and is replicated three times among the six rings. The subplot factors of species identity and N treatment were randomly assigned and replicated in individual plots among the six rings. CO2 was added in elevated treatments during all daylight hours from April 9 to October 16, 1998, and from April 20 to November 9, 1999. Tests found no direct effect of elevated CO2 on dark respiration (Tjoelker et al., 2001). During CO2 enrichment periods, 1-min averages were within 10% of the target concentration 94% of the time in 1998 and 95% of the time in 1999. Beginning in 1998, the plots assigned to the high N treatment were amended with 4 g N m−2 yr−1, applied over three dates each year, while the low N treatment soil was unamended. During the two growing seasons of treatments during this study, no severe dry periods occurred.

The 16 perennial species used in this study were all native or naturalized to the Cedar Creek Natural History Area. They include four C4 grasses (Andropogongerardii Vitman, Boutelouagracilis, Schizachyrium scoparium (Michaux) Nash, Sorghastrum nutans (L.)(Nash), four C3 grasses (Agropyron repens (L.) Beauv., Bromus inermis Leysser, Koeleria cristata Pers, Poa pratensis L.), four N-fixing legumes (Amorpha canescens Pursh, Lespedeza capitata Michaux, Lupinusperennis L., Petalostemum villosum Nutt.) and four nonN-fixing herbaceous species (Achilleamillefolium L., Anemone cylindrica A. Gray, Asclepias tuberosa L., Solidago rigida L.). Species hereafter are referred to by their genus. Monocultures of all species were replicated twice at all four combinations of CO2 and N levels. Plots were regularly weeded to remove unwanted species. In June and August of 1998 and 1999 we assessed above- and belowground (0–20 cm) biomass and soil solution N concentrations (extracted using 0.01 mol KCl). For most analyses in this paper we use the mean values per plot from these four harvests. A 10 × 100 cm strip was clipped at just above the soil surface, all matter was collected, sorted to live material and senesced litter, dried and weighed. Roots were sampled at 0–20 cm depth using three 5-cm cores in the area used for the aboveground biomass clipping. Roots were washed, sorted into fine (< 1 mm diameter) and coarse classes and crowns, dried and weighed. Volumetric soil moisture levels (0–20 cm depth) were assessed periodically (on 18 sampling periods) in all plots in 1998 and 1999 using time-domain reflectometry (Baker, 1990). A composite sample was taken from aboveground and belowground biomass from each plot from the August harvests of each year, ground and analysed for N using a CHN analyser (Carlo-Erba Strumatzione, Milan, Italy). To estimate total plant N stocks we multiplied whole plant percentage N (averaged across years) by the mean whole plot biomass (averaged from all harvests).

Data analysis

ANOVA was used in two complementary ways, including either species or functional group as a treatment effect. In ANOVA all treatment effects were considered fixed. Using F-tests, the effect of CO2 (1 df) was tested against the random effect of ring nested within CO2 (4 df). The main effects of functional group (3 df) was tested against the random effect of species nested within functional group (12 df). The main effects of species (15 df), and N (1 df), and interactions between CO2, N, and either functional group or species were tested against the residual error. Additionally, to explicitly test hypotheses about preplanned specific functional group contrasts (e.g. C3 vs C4), their effects were partitioned into single-degree-of-freedom contrasts. We also evaluated the proportional distribution of biomass aboveground vs belowground. Since this is a plot-scale measure for assemblages that had been developing in the field for three seasons it should not be taken as a measure of allocation. Instead, it reflects the balance between allocation and turnover (Reich, 2001).We also used separate and same slopes regression to test for slope and intercept differences among treatments in the relationships between soil resources (e.g. % soil water or N) and fine root biomass. We used repeated measures ANOVA to test whether responses to CO2 or N varied among times of year (June vs August), among years, or among all four harvests. There were no interactions between treatments and time, hence all results are presented averaged across harvests and years. All statistical analyses were made using JMP 4.0.1 software.


Functional groups differed significantly in shoot, root and total biomass (Tables 1, 2; Fig. 1), with C3 grass monocultures highest and legumes lowest. For total biomass, averaged across all species or functional groups, there were marginally significant main effects of CO2 (+11%, mean of enhancement) and N (+7%) (Tables 2, 3; Fig. 1) and there were no significant CO2 × N interactions (Tables 2, 3). In fact, there were no significant CO2 × N interactions, or significant three–way interactions involving CO2 and N, for any of the measured variables in this study (refuting H5), and hence responses to CO2 and N will be presented separately, pooled across levels of the other variable.

Table 1.  Means of belowground and aboveground biomass of species under contrasting CO2 and N treatments for 128 plots. Values shown are for 4 plots per species at each CO2 or N level, pooled across treatments otherwise and averaged over four harvests in two years. The mean standard errors of the adjusted least squares means (LSM) for belowground and aboveground biomass, respectively, were 59 and 21 g m−2 for species and 103 and 14 for functional groups
  Belowground biomass (g m−2)Aboveground biomass (g m−2)
Functional groupSpeciesAmbient CO2Elevated CO2Low NHigh NAmbient CO2Elevated CO2Low NHigh N
C4 grassAndropiogon gerardii575  525543  557261283263282
 Bouteloua gracilis472  562471  564253200209244
 Schizachyrium scoparius359  349293  415177153133196
 Sorghastrum nutans569  513614  468248221199270
 Mean493  489480  502236215204247
C3 grassAgropyron repens80611148831037308307282333
 Bromus inermis828  755750  834306281253333
 Koeleria cristata645  714666  692253225194284
 Poa pratensis99110589291120168200123245
 Mean817  910807  920256252210299
ForbAchillea millefolium7001014944  770268293266295
 Anemone cylindrica183  288324  147  36  53  60  29
 Asclepias tuberosa  97  172  91  179  11  14  12  13
 Solidago rigida597  690706  581315386328372
 Mean395  541516  419160185167179
LegumeAmorpha canescens172  160208  124  59  26  59  26
 Lespedeza capitata274  407315  366  67153143  77
 Lupinus perennis216  303270  250279383339323
 Petalostemum villosum135  122118  140142  39  27154
 Mean200  246226  220137151143145
Table 2.  ANOVA summary for biomass, N, and soil water measures in 128 plots
 CO2 × N × Functional group analyses
ParameterR2CO2NGroupCO2× GroupCO2× (C3 vs C4)CO2 × C3 N-fixer vs C3 nonfixerN x groupN vs N-fixer vs nonfixer
  1. R2 shown for whole model. P-values shown for main effects and interactions (bolded when P < 0.10). Interactions involving CO2 × N were not significant at P < 0.10) and hence are not shown. Group, functional group contrasts of the four groups; C3 vs C4, C4 grasses vs all other C3 species; C3 N-fixer vs C3 nonfixer, legumes vs forbs plus C3 grasses; N-fixer vs nonfixer, legumes vs all other nonfixer species.

Total biomass0.880.05    0.0020.040.001    0.08
Belowground biomass0.870.08    0.770.0060.14    0.0050.020.007    0.67
Aboveground biomass0.840.47    0.00030.510.30    0.140.770.01    0.0006
Root fraction0.650.40    0.020.390.70    0.310.720.92    0.36
Total plant N pool0.800.27< 0.00010.180.51    0.180.650.0006< 0.0001
Belowground percentageN0.850.30    0.030.0020.38    0.0007
Aboveground percentageN0.890.03    0.0050.070.02< 0.00010.080.002    0.0002
% soil water0.700.60    0.00030.110.11    0.650.110.009    0.15
Soil nitrate concentration0.690.08    0.0010.0020.03    0.47
Soil N concentration0.740.22     0.00050.0020.28     0.070.710.03    0.84
Figure 1.

Total biomass (aboveground plus belowground, 0–20 cm depth) (g m−2) of 16 species as affected by elevated vs ambient CO2 treatments (pooled across N treatments) and by high N (addition of 4 g N m−2 yr−1) vs unamended soil treatments (pooled across CO2 treatments). Top panel: grey columns, elevated CO2; black columns, ambient CO2. Bottom panel: grey columns, low N; white columns, high N. Each value is a mean of four harvests (June and August in each of 1998 and 1999) for 4 plots per species-treatment combination. The mean standard errors of the adjusted least squares means (LSMs) were 66 g m−2 for species. Statistical details provided in Tables 2 and 3.

Table 3.  ANOVA summary for biomass, N, and soil water measures in 128 plots
 CO2 × N × Species analyses   
ParameterR2CO2NSpeciesCO2 x SpeciesN x Species
  1. R2 shown for whole model. P-values shown for main effects and interactions (bolded when P < 0.10), except those involving CO2 × N which were not significant at P < 0.10.

Total biomass0.940.06    0.07< 0.00010.01    0.01
Belowground biomass0.930.11    0.73< 0.00010.04    0.02
Aboveground biomass0.940.53< 0.0001< 0.00010.0009    0.0004
Root fraction0.870.33    0.004< 0.00010.02    0.003
Total plant N pool0.910.29< 0.0001< 0.00010.02    0.0003
Belowground percentageN0.910.34    0.04< 0.00010.98    0.05
Aboveground percentageN0.960.01    0.001< 0.00010.01< 0.0001
% soil water0.820.58    0.0005< 0.00010.68    0.13
Soil nitrate concentration0.860.09    0.0004< 0.00010.002    0.12
Soil N concentration0.870.15     0.00030.00010.01    0.29

Biomass response to elevated CO2

There were marked differences among functional groups in terms of biomass response to CO2 (Tables 1, 2; Figs 1, 2). Forbs, legumes, and C3 grasses increased total biomass by 31%, 18%, and 9%, respectively, under elevated CO2 whereas C4 grass monocultures had 3% lower total biomass. The interaction term was significant (P < 0.10) for the CO2–group interaction and more so (P < 0.002) for the C3 vs C4 contrast (Table 2), indicating that the C4 grasses did have less enhancement of biomass than C3 species in general, supporting (H1). There was no evidence that the N-fixing legumes responded more positively to elevated CO2 than the nonfixing C3 species (Figs 1, 2; Tables 1, 2), refuting (H2). The effects of CO2 on biomass, and differences in response among functional groups, were largely manifest belowground (Tables 1, 2).

Figure 2.

Proportional response of total biomass, plant N pool, percentage soil water, and soil solution N concentration to CO2 and N treatments (otherwise pooled) on average for species in four functional groups. See Table 2 for statistics.

There was substantial variation in response to elevated CO2 among species, shown by significant species–CO2 interactions for shoot, root and total biomass (Tables 1, 3). Seven species increased total biomass under elevated CO2 by at least 100 g m−2, including three forbs, two C3 grasses and two legumes (Fig. 1). The largest responses to elevated CO2 were shown by a forb, Achillea (+339 g m−2, +35%), a C3 grass, Agropyron (+307 g m−2, +28%) and two legumes, Lespedeza (+218 g m−2, +64%) and Lupinus (+190 g m−2, + 38%), all of which were statistically significant (P < 0.05) using post hoc tests. However, some species within all three of these C3 functional groups also had modest or minimal responses to elevated CO2 (Fig. 1).

Biomass response to N addition

Functional groups responded differently (P < 0.01) to N treatment in terms of both aboveground and belowground biomass (Tables 1, 2). At high N, C3 grasses showed the greatest increase in total biomass of all functional groups (Tables 1, 2; Figs 1, 2), supporting (H3), and the legumes responded less positively in general than the nonfixers, supporting (H4). However, the forb group also failed to respond to high N with increased biomass production. Moreover, except within the C3 grass group, variation in biomass response to N among species within functional groups was substantial, shown by (Table 1, Fig. 1).

There were significant species–N interactions for shoot, root and total biomass (Tables 1, 3; Fig. 1). Seven species increased total biomass under high N by at least 140 g m−2, including all four C3 grasses, two C4 grasses, and a legume, with Poa (+313 g m−2, +30%), Agropyron (+206 g m−2, +18%) and Schizachyrium (+163 g m−2, +37%) showing the largest increases.

Variation among functional groups and species

How do differences among functional groups compare with differences among species? We assessed the coefficient of variation (CV) of functional group and species means, and of the responses to treatments, for many variables. Results for total biomass are representative of the general trends. The CV among groups for total biomass at each of the four CO2 and N treatment combinations was slightly less than the CV for all species assessed collectively (Table 4). However, the CV among species within groups varied extremely among groups, from as low as 15–20% for C3 grasses to as high as 70–80% for the forbs. The response of biomass to CO2 also had a lower CV among groups (102%) than among all species (187%).

Table 4.  Coefficient of variation of total biomass per plot among species within functional groups, among all species, and among functional groups, for four treatment combinations of CO2 and N. Mean values per species or functional group based on averages of all plots in all harvests
 Low NHigh N
 Ambient CO2Elevated CO2Ambient CO2Elevated CO2
C3 grass spp.11.512.612.621.8
C4 grass spp.25.724.412.020.1
Forb spp.85.370.476.483.7
Legume spp.43.370.341.062.8
All species53.851.455.158.5
Functional groups39.936.852.746.7

Proportional biomass distribution under elevated CO2 and N treatments

Functional groups did not differ significantly in the proportion of total biomass found belowground, that is root fraction (Tables 2, 5). The C4 grasses had root fractions between 67% and 69%, the C3 grasses had root fractions between 74% and 86%, and species within other functional groups varied widely. Among all species, Lupinus had the lowest root fraction (43%) but other legumes had intermediate or high root fractions (69–82%).

Root fractions under elevated CO2 (0.74 on average) were not significantly different than under ambient CO2 (0.72). Root fraction was lower (P < 0.005) in the high N (0.71) than the low N treatment (0.75). These results support H6. Root fraction was unrelated to total biomass. There were no significant functional group × N or group–CO2 interactions for root fraction (Table 2), but there were significant species × CO2 and species–N interactions (Table 3). Thus, functional groups did not show different biomass distribution response to CO2 or N, but species did. Six species, representing all four functional groups, had lower root fraction under high N (Table 5). Petalostemum had a root fraction of 0.84 under low N and 0.54 under high N, while another legume species (Lespedeza) had greater root fraction under high N (0.81) than low N (0.72). Four species, one from each functional group, had higher root fraction under elevated CO2 (Table 5). Petalostemum was again the most plastic in root fraction, showing root fraction of 0.55 in ambient CO2 and 0.82 under elevated CO2. Again, Lespedeza showed an opposite response to most other species, having lower root fraction under elevated than ambient CO2. Thus, species were quite varied in the magnitude of biomass distribution responses, with no relation to functional group membership, and in fact the two species with the most pronounced and opposite behaviour were both legumes.

Table 5.  Means of root fraction (root biomass/total biomass) of species under contrasting CO2 and N treatments for 128 plots
  Root fraction (root dm/total plant dm)
Functional groupSpeciesAmbient CO2Elevated CO2Ambient NEnriched N
  1. Values shown are for 4 plots per species at each CO2 or N level, pooled across treatments otherwise and averaged over four harvests in two years. The mean standard errors of the adjusted LSMs were 0.04 for species and 0.06 for functional groups.

C4 grassAndropogon gerardii0.690.650.670.67
 Bouteloua gracilis0.650.740.690.69
 Schizachyrium scoparius0.680.690.700.68
 Sorghastrum nutans0.690.690.750.63
C3 grassAgropyron repens0.730.790.760.75
 Bromus inermis0.740.730.760.71
 Koeleria cristata0.720.760.770.71
 Poa pratensis0.860.840.890.82
ForbAchillea millefolium0.710.770.770.72
 Anemone cylindrica0.840.850.830.86
 Asclepias tuberosa0.910.930.910.92
 Solidago rigida0.640.640.680.61
LegumeAmorpha canescens0.810.850.820.83
 Lespedeza capitata0.800.730.720.81
 Lupinus perennis0.430.430.430.43
 Petalostemum villosum0.550.820.840.54

Total plant N and tissue N concentration

Averaged across all functional groups or species, the total plant N pool was significantly higher with N addition and unchanged with elevated CO2 (Tables 2, 3). Thus, despite greater root biomass under elevated CO2, total cumulative N uptake did not increase, and hence there was substantial dilution of N in biomass (see below). Species differed markedly (P < 0.001) in how plant N pools responded to CO2 treatments, however (Table 3, Fig. 3). Lespedeza and Lupinus had large increases in total plant N (of roughly 40–60%, +3.3–4.0 g N m−2) with increased CO2, whereas most other species had similar total N in the elevated vs ambient CO2 treatment. Response of total plant N pools to N treatment differed (P < 0.05) by functional groups; with forbs showing minimal increase, whereas both grass groups and legumes showed substantial increase in total N (Table 2; Figs 2, 3).

Figure 3.

Means of total plant N pools (g m−2) of 16 species under contrasting CO2 and N treatments. Top panel: grey columns, elevated CO2; black columns, ambient CO2. Bottom panel: grey columns, low N; white columns, high N. The mean standard errors of the adjusted LSMs were 0.8 g m−2 for species. See Tables 2 and 3 for statistics.

Functional groups differed in tissue percentage N (Tables 2, 6). Root and shoot percentage N responded to treatments differently, in that roots had less pronounced treatment effects and there were fewer interactions between functional groups or species and treatments (Tables 2, 3, 6). For aboveground tissue, CO2 treatment reduced percentage N by 11% on average and the high N treatment increased percentage N by 8% on average (pooled across all other sources of variation). Both functional groups and species differed significantly (Tables 2, 3) in response of aboveground percentage N to both treatments. The C4 grasses did not have lower percentage N under elevated CO2 (Table 6). All C3 functional groups tended to have substantially lower tissue percentage N under elevated CO2 (Table 6), but species varied markedly in this respect. Agropyron, Amorpha, Anemone, Asclepias, Lespedeza, and Poa (equally representing the three C3 groups) decreased percentage N by the greatest amounts under elevated CO2, and most other C3 species had smaller or negligible decreases. Except for the legumes and Asclepias, all other species had modest or marked increases in aboveground percentage N with N addition (Table 6).

Table 6.  Means of tissue percentage N of species under contrasting CO2 and N treatments for 128 plots. Values shown are for 4 plots per species at each CO2 or N level, pooled across treatments otherwise and averaged over August harvests in 1998 and 1999
  Aboveground % NBelowground % NTotal % N
Functional groupSpeciesAmb CO2Elev CO2Low NHigh NAmb CO2Elev CO2Low NHigh NAmb CO2Elev CO2Low NHigh N
  1. The mean standard error of the adjusted LSMs was 0.10, 0.09, and 0.08% for aboveground, belowground and total percentageN for species and 0.14, 0.07, and 0.07% for aboveground, belowground and total percentageN for functional groups.

C4 grassAndropogon gerardi1.091.150.931.310.951.000.881.070.981.050.901.13
 Bouteloua gracilis1.411.331.111.631.141.070.941.
 Schizachyrium scoparius1.
 Sorghastrum nutans0.910.890.801.000.880.870.820.940.890.890.810.97
C3 grassAgropyron repens1.260.890.961.190.840.720.720.840.940.730.750.92
 Bromus inermis0.911.000.801.100.910.850.870.890.910.890.850.94
 Koeleria cristata1.431.301.201.530.951.031.010.971.
 Poa pratensis1.421.101.171.370.950.860.890.911.030.910.931.01
ForbAchillea millefolium1.471.271.351.390.850.710.780.781.060.870.931.00
 Anemone cylindrica1.891.551.571.881.371.201.191.381.491.301.271.51
 Asclepias tuberosa3.442.943.512.861.451.421.491.381.901.651.771.78
 Solidago rigida1.070.800.801.081.090.840.791.141.100.790.801.10
LegumeAmorpha canescens2.702.172.472.401.481.441.401.531.731.551.581.71
 Lespedeza capitata2.351.952.122.181.891.921.981.832.001.932.021.91
 Lupinus perennis1.521.431.431.522.
 Petalostemum villosum2.262.252.462.051.351.331.331.351.721.511.611.62

Soil water

On average, plots had lower percentage soil water (% SW) in the high N than low N treatments, as hypothesized (H7) (Tables 2, 3, 7; Fig. 2). Moreover, species and functional groups differed marginally in percentage SW and in how percentage SW changed with CO2 and N. C3 grasses had the lowest percentage SW, C4 grass and forbs intermediate and legumes the highest percentage SW (Table 7), likely reflecting the influence of differential root biomass on water uptake. To sum, groups are different in percentage SW, but apparently mostly because they have different fine root biomass. Among species, however, differences in percentage SW do not neatly follow differences in total or root biomass-Koeleria (a C3 grass) had the lowest percentage SW and Lupinus, Agropyron and several other species the highest.

Table 7.  Means of percentage soil water and soil solution N concentrations for species under contrasting CO2 and N treatments for 128 plots. Values shown are for 4 plots per species at each CO2 or N level, pooled across treatments otherwise and averaged over multiple samplings for soil water, and two harvests for soil N, in each of 1998 and 1999. The mean standard error of the adjusted LSMs for percentage soil water was 0.03% for species and 0.2% for functional groups, and 0.1 mg kg−1 for soil N for both species and functional groups
   % soil waterSoil solution N conc. (mg/kg)
Functional groupSpeciesAmb CO2Elev CO2Low NHigh NAmb CO2Elev CO2Low NHigh N
C4 grassAndropogon gerardii7.
 Bouteloua gracilis6.
 Schizachyrium scoparius7.
 Sorghastrum nutans7.
C3 grassAgropyron repens7.
 Bromus inermis6.
 Koeleria cristata6.
 Poa pratensis6.
ForbAchillea millefolium7.
 Anemone cylindrica7.
 Asclepias tuberosa7.
 Solidago rigida6.
LegumeAmorpha canescens8.
 Lespedeza capitata7.
 Lupinus perennis8.
 Petalostemum villosum7.

On average, C3 grasses showed the greatest increase in percentage SW with elevated CO2 and the greatest decrease in percentage SW at high N (Table 7; Fig. 2). Except for the legumes, all species had slightly or substantially higher percentage SW under elevated than ambient CO2 (Table 7), even when root and total biomass were higher in the latter treatment (e.g. Achillea, Agropyron, Anemone, Bouteloua, Poa) (supporting H7b). Bromus had the most pronounced (P = 0.05) increase in percentage SW under elevated vs ambient CO2, consistent with its slightly lower biomass under elevated CO2. For N treatments, Agropyron and Koeleria (C3 grasses) both had substantially greater depression of percentage SW (both P < 0.05) in high N than other species.

Even accounting for variation in fine root biomass, there were significant differences in percentage SW due to CO2 and N treatments. When data for all plots were pooled and percentage SW was plotted against fine root biomass, there was significantly higher percentage SW under elevated than ambient CO2 and lower percentage SW under elevated than ambient N (Fig. 4a,b).

Figure 4.

(a,b) Mean percentage volumetric soil water (averaged over 18 sampling periods over two years) in relation to mean fine root biomass (averaged over 4 harvests in 2 yr) for 16 species monocultures under contrasting CO2 and N treatments (pooled across the other treatment). Open circles, ambient CO2; closed circles, elevated CO2; open triangles, low N; closed triangles, high N. The slopes were not significantly different among treatments in either case, but the elevation of the line was significantly (P < 0.025) different in each case. (c,d) Mean soil solution N concentration (mg/kg) in relation to mean fine root biomass (both averaged over 4 harvests in 2 yr) for 16 species monocultures under contrasting CO2 and N treatments (pooled across the other treatment). Open circles, ambient CO2; closed circles, elevated CO2; open triangles, low N; closed triangles, high N. The slopes were not significantly different among treatments in either case, and the elevation of the line was significantly (P < 0.025) different for contrasting N treatments but not contrasting CO2 treatments.

Soil solution N

The high N treatment increased soil solution nitrate and total N by roughly 40–50% on average (Table 7; Fig. 2). In contrast, and supporting H7a, elevated CO2 reduced soil nitrate and total N by roughly 20–25% (Table 7; Fig. 2), consistent with increased fine root biomass under elevated CO2. Functional groups differed in soil solution N, with legumes having the highest and C3 grasses the lowest levels (Tables 2, 7). These functional group differences were roughly opposite to patterns of fine root biomass. Functional groups differed in the extent of depletion of soil solution N due to CO2, with the C4 group having little effect compared to the C3 groups in general (Tables 5, 7). For N treatment, legumes had a significantly greater increase in soil solution N than the C3 grass species, likely due to the degree of soil solution N depletion associated with differences in fine root biomass.

Accounting for variation in fine root biomass, there were significant differences in soil N concentrations due to N treatments, but not to CO2. When data for all plots were pooled and soil solution N was plotted against fine root biomass, plots had significantly higher soil N under high than low N but did not differ under the contrasting CO2 treatments (Fig. 4c,d).

For nonlegumes, there was an inverse relationship between total plant N pools and soil solution N concentrations (P < 0.001), likely because vegetation on plots with higher fine root biomass take up N (which is incorporated into plant tissues) while driving down the soil solution N concentration (Fig. 5). At any soil solution N pool, the N deposition treatment had higher plant N pools. Two of the legumes (Petalostemum and Amorpha) had slightly higher soil solution N pools at any given plant N pool but fit more or less within the general scatterplot relationship for the non-legumes (P < 0.001, r2 = 0.44 for the 56 plots within each N treatment level) (Fig. 5). By contrast, the other two legumes, Lespedeza and Lupinus, always had higher soil solution N pools at any given plant N pool and plots for these two species fell far from the relationship for the other 14 species (Fig. 5).

Figure 5.

Mean soil solution N concentration (mg/kg) in relation to mean total plant N (g m−2), for plots under low and high N treatments, pooled across CO2 treatments. The curves shown are for all nonN-fixers plus Amorpha and Petalostemeum, and the relationship between log10 soil N and log10 plant N was significant (P < 0.001, R2 = 0.44) in both cases. Closed circles, 14 species; open circles, Lespedeza and Lupinus.


A surprising number of species in this study had modest or negligible increases in biomass production in response to CO2 or N fertilization (Fig. 1). Our major hypotheses regarding functional group responses to CO2 and N were neither consistently supported nor rejected, and will be discussed below. Overall, these grassland species were less responsive to elevated CO2 (see also Lee et al., 2001) than has been generally found elsewhere, although some studies have found weak responses to elevated CO2 (Koch & Mooney, 1996; Körner & Bazzaz, 1996; Curtis & Wang, 1998; Poorter, 1998). The mean increase due to elevated CO2 in total biomass of all 12 C3 species in our study was 16%, considerably less than the 29%, 42% and 44% increases, respectively, reported in reviews of C3 woody plants (Curtis & Wang, 1998), C3 grasses in general (Poorter, 1993), and C3 Poaceae (Wand et al., 1999). The mean response of biomass to elevated CO2 for the C4 grasses in our study (−3%) is also less than the mean increases of 22% and 33% reported for C4 grasses in general (Poorter, 1993) and for C4 Poaceae (Wand et al., 1999), respectively. Possible explanations for the limited CO2 response in our study compared with previous studies include the relatively low fertility of our site, the general high level of adaptation to infertility among our study species, and the moist conditions during the study years. Alternatively, only a small fraction of the studies used in the reviews cited above were grown under realistic field conditions, and even fewer without chambers. It is important to note that a high degree of photosynthetic acclimation among these species (Lee et al., 2001) can broadly explain the small growth response to elevated CO2 at our site, though the correlation between individual species photosynthetic vs. biomass responses was weak (P. B. Reich et al., unpublished). Testing whether long-term responses to elevated CO2 of grassland species in the field are generally less than those of shorter-term and more controlled experiments will require a larger data base than is currently available.

Functional group and species responses to elevated CO2

As a group, C4 grasses did show significantly less biomass enhancement in response to elevated CO2 than the C3 functional groups (supporting H1) and no C4 grass species showed a significant increase. These biomass responses are consistent with leaf level photosynthesis data; during the 2-yr study period, the C4 grass species showed no enhancement of photosynthesis under elevated CO2, in contrast to a modest increase on average for the C3 species (Lee et al., 2001).

Two of the four legumes in our study, Lupinus and Lespedeza, showed marked production increases under elevated CO2 but the other two did not. Thus (H2) – a hypothesis based on the general tendency of N-fixers to respond more to elevated CO2 than nonfixers (Soussana & Hartwig, 1996; Clark et al., 1997; Hebeisen et al., 1997; Lüscher et al., 1997, 1998; Schenk et al., 1997) – apparently did not hold. Can we explain this marked difference in response among legume species?

Although we know little about variation in the degree of N-fixation among these four legumes, we can make some indirect inferences based on the data from this study. Lupinus and Lespedeza plots had somewhat higher soil solution N pools and two to three times as much plant N per plot as Amorpha and Petalostemum, which we take as evidence of greater N fixation in the former two species, since all else being equal, with greater root biomass they should have otherwise depleted the soil solution N to a level lower than the other two legumes. Moreover, for all nonfixers plus Amorpha and Petalostemum pooled, there was a significant inverse relationship (Fig. 5) between soil solution N and plant N pools, consistent with earlier studies relating differences among species in root biomass to depletion of soil solution N pools in grasslands at Cedar Creek (Tilman & Wedin, 1991). Lupinus and Lespedeza did not follow this inverse relationship and had much higher soil solution N for a given plant N pool, suggesting relatively high rates of N fixation.

In addition, for the 12 nonfixer species (plus Amorpha and Petalostemum) the plant N pool was relatively unchanged under elevated CO2. This suggests that increased biomass in response to elevated CO2 by itself did not result in greater total plant N acquisition in nonfixers. In contrast, Lupinus and Lespedeza had marked increases in plant N pool in elevated CO2 (4.0 and 3.3 g m−2, respectively). Hence, it is plausible that Lupinus and Lespedeza more vigorously fix N than the other legumes, and responded to elevated CO2 by increasing their own N supply, leading to heightened uptake of both C and N. In support of this idea, using the 15N isotope dilution method for Lupinus monocultures, the proportion of N derived from fixation increased by 38% under elevated CO2 (T. D. Lee et al., unpublished). Thus, based on indirect evidence from this study, the legumes with greater N-fixation tendencies were more responsive to elevated CO2 than those with lesser N-fixation tendencies, supporting the concept of N-fixation enhancing CO2 responses (H2), but suggesting that heterogeneity among N-fixers may limit the generalizability of this idea.

Functional group and species responses to increased N supply

Our hypotheses about biomass responses to increased N supply (H3 to H5) were also supported with mixed results. The C3 grasses increased biomass the most at high N and the legumes as a group responded little to N addition, as predicted (H3, H4). However, one of the four legume species, Petalostemum, did increase biomass at high N, and perhaps related, this species did not respond positively to elevated CO2 and may have had a low rate of N fixation. All eight grasses, regardless of photosynthetic pathway, had large increases in aboveground biomass at high N, and 7 of the 8 also had large increases belowground. However, surprisingly, only one of four forbs responded positively to N addition in terms of total biomass. Perhaps these species are poor competitors for soil N in relationship to microbes? If that was so (i.e. microbial uptake depleted the available soil N), soil solution N should be low, yet it was higher for the forbs than for the grasses. Moreover, two of the three forbs which did not increase biomass at high N did have higher tissue percentage N. The failure of these species to respond positively with increased production under high N is difficult to explain.

Finally, there were no CO2 × N interactions and hence no tendency for any functional group to respond differently to CO2 as a function of N supply (H5). Elevated N supply did not increase monoculture response to CO2, unlike a number of other studies (e.g. Owensby et al., 1994; Zanetti et al., 1996; Curtis & Wang, 1998; Zak et al., 2000). Whether this is due to the relatively smaller N addition in this study (4 g N m−2 yr−1) than in some others (e.g. N treatments were as high as 56 g N m−2 yr−1 and more closely mimic agricultural N addition rates, Zanetti et al., 1996) can not be answered without a larger number of studies of elevated CO2 effects under different N addition regimes.

Controls on percentage soil water

How can we explain CO2 and N treatment effects on percentage soil water? Biomass per plot had a large influence on percentage SW (Fig. 4), and both species and treatments contributed to variation in biomass, although the former was dominant. For N, even for a given fine root mass, plots under high N had lower percentage SW. There was no effect of N on leaf diffusive conductance (Lee et al., 2001) (and hence on leaf-level water loss), but high N did lead to a lower root fraction (Tables 3, 5). Hence, high N plots support a higher aboveground biomass and likely a higher leaf area index and transpirational surface, perhaps explaining the lower percentage SW under high N at a given fine root biomass.

By contrast, elevated CO2 leads to greater biomass which, all else being equal, should lead to greater depletion of soil water. However, individual species actually had lower percentage SW under elevated CO2 despite greater biomass (Table 7) and overall, percentage SW at any given fine root biomass was higher under elevated CO2. Elevated CO2 did not affect root fraction, so a shift in absorbing (roots) vs transpirational surfaces (canopy) can not be invoked as an explanatory factor. In a companion study (Lee et al., 2001) we found a consistent and roughly 25% decrease in leaf conductance across species, functional groups and years. An analysis (data not shown) that includes functional group and leaf conductance found both factors significantly (P < 0.05) associated with percentage SW, with percentage SW negatively related to conductance. Hence, reduced conductance contributes to the tendency for elevated CO2 plots to have lower percentage SW.

Other functional traits and groupings

Variation in a number of other functional traits, including intrinsic growth rate, leaf gas exchange capacity, resource depletion capacity, sink strength, plant strategy, and root symbiont status, has been proposed to explain differential responsiveness to elevated CO2, added N and a variety of other global change factors (Reich, 1987; Hunt et al., 1991, 1993; Díaz, 1995; Poorter et al., 1996; Wedin & Tilman, 1996). Such explanations have met with some success with other factors, but less so vis-à-vis CO2 (Poorter et al., 1996; Volin & Reich, 1996). None of these hypotheses moreover, appear capable of fully explaining differences in responses among the 16 species in this study. The species that were most productive under ambient CO2 and soil N conditions were not consistently responsive to either elevated CO2 or N, nor were the less productive species consistently responsive. Similarly, for 10 of the 16 species we also measured relative growth rate of seedlings under controlled conditions (P. B. Reich et al., unpublished) and found no relationship between proportional biomass enhancement due to CO2 in field plots and individual seedling RGR. Species which reduce the soil N concentration under ambient conditions did generally respond more positively to N enrichment, but with two important exceptions (Achillea and Solidago).

Among the nonlegumes, tissue percentage N in the low N treatment did not predict which species would respond to N enrichment (data not shown), but it did predict the relative responsiveness to elevated CO2 (Fig. 6). For C3 and C4 species separately, species with greater aboveground tissue percentage N had a greater increase in biomass in response to elevated CO2 than those with lower percentage N. The slopes were similar for both groups, and the relationship was at a lower ‘elevation’ (i.e. intercept was lower) for the C4 than C3 species. Why do these patterns occur? A plausible hypothesis follows from physiological principles. Species with higher tissue percentage N have a higher carboxylation capacity (Evans, 1989) such that if they were ‘operating’ on the steep part of the A-Ci curve (Farquhar & Sharkey, 1982) they should have a greater enhancement due to elevated CO2 than species with lower tissue percentage N (this could explain patterns within each group). Moreover, since C4 plants are already near saturated at ambient CO2 (i.e. on a shallower part of the curve) they are less responsive as a group to elevated CO2, even for a given leaf percentage N (this could explain the lower position of the overall line for this group). With their different C : N dynamics, legumes would not necessarily be expected to follow similar patterns. Although many processes beyond leaf level photosynthesis influence ecosystem scale biomass enhancement to elevated CO2 (such as canopy architecture, turnover rates, tissue morphology, biomass distribution, and phenology), leaf-level processes provide the starting point for carbon acquisition which leads to biomass accumulation, and which drives increased biomass accumulation under elevated CO2. Whether this is a robust general relationship requires further testing in other common garden experiments in the field.

Figure 6.

The percentage enhancement of total biomass due to elevated CO2 for all nonN-fixer species, in relation to the aboveground tissue percentage N of plants under ambient CO2 conditions. Open circles, C4 species; closed circles, C3 species. The regression relationships were significant (P < 0.001) for C3 and C4 species considered separately (R2 = 0.72 and 0.99, respectively), with a similar slope (P > 0.10) but different intercept (P < 0.05).

Implications and conclusions

There have been many isolated potted plant studies of elevated CO2 responses. Studies of monocultures in the field with free-air CO2 enrichment provide information that is closer to natural conditions. Moreover, if we hope to be able to use functional groupings to generalize about, and quantitatively model responses to elevated CO2 and N of mixed species communities, then we need to understand variation in species and functional group responses under simpler monoculture conditions. If responses to elevated CO2 and N deposition are largely a function of changing resource supply then growth in mixed communities is complicated by species interactions that can change as a result of these agents (Owensby et al., 1993, 1994). Competition often involves growing with other species that differentially utilize resources (e.g. soil solution N, Tilman & Wedin, 1991) or supply them (e.g. N-fixers) by dint of species ecophysiological differences. Hence, understanding species responses to enriched CO2 or N in mixed communities is complicated because competition ensures that any given species does not necessarily have access to increased supplies of a resource that is added to an ecosystem. Nonetheless, responses of species mixtures to elevated CO2 may be related to their responses in monocultures (Navas et al., 1999). Therefore, interpretation and evaluation of species under interspecific competition (e.g. Warwick et al., 1998; Leadley et al., 1999) will be aided by a better understanding of their responses growing under intraspecific competition, as in the monocultures of this study.

Our results are only somewhat encouraging vis-à-vis the use of functional groups. Functional groups did often respond significantly differently to CO2 or N, and hence do provide some meaningful information without knowledge about individual species. However, there was also substantial variation in response among species within groups. Alternative classifications based on measured continuous traits were generally no more useful than the a priori defined functional groups. These results suggest that current trait-based functional classifications may be useful, but not sufficient for understanding plant and ecosystem responses to elevated CO2 and N deposition.


We acknowledge funding from the U.S. Department of Energy and the National Science Foundation Cedar Creek Natural History Area, Long-Term Ecological Research site in support of this research. We thank G. Hendrey, K. Lewin, C. Lehman, K. Wrage, S. Jose and many undergraduate interns whose work contributed to myriad phases of this research.